"AI consultant" is a job title that tells you almost nothing. Here is the concrete version, written for an owner-run business in the Town of Huntington with somewhere between five and fifty people on the payroll.
rTech works out of Melville, which sits inside the Town of Huntington, and is a member of the Melville Chamber of Commerce. That matters less than the practical part: a twenty-minute drive covers the village, Melville, and the length of Route 110, and standing in the room where the work happens is usually worth more than three video calls.
What an AI consultant does, stated plainly: pick one recurring workflow, watch it run, write down how it actually works, name the points where a person has to decide, and identify which mundane parts a machine can carry. That is the whole job. Everything else is either the follow-through or a sales pitch wearing the same title.
What it is not
It is not a software purchase. Recommendations are vendor-agnostic and run on whatever stack is already in the building: Claude, ChatGPT, Copilot, or something else. rTech does not host client AI and does not hold anyone's keys or billing.
It is also not a headcount conversation. The point of moving the mundane parts off a plate is that the person whose plate it was can carry more of the work they were hired to do. A system that takes away someone's judgment has taken away the person, and that is not the trade on offer here.
Three Huntington situations
These are the three shapes the phone call usually takes around here. None of them are clients, and none of them carry anybody's confidential material: they are patterns, described so you can tell whether one of them is yours.
A shop on New York Avenue
A village storefront with six people. Every supplier email, every special order, and every "can you hold this for me" arrives in one inbox that two people half-watch. Nothing is lost often, but something is lost occasionally, and the owner is the backstop for all of it.
The work here is not a chatbot on the website. It is a drafting and sorting step: incoming messages read into a short list of what is open, what is waiting on a supplier, and what somebody promised a customer, with a draft reply prepared for the ones that are routine. What the team keeps doing: deciding what to order, what to hold, what to comp, and what to say to the customer standing in front of them. What gets lifted off: the reading, the sorting, the first draft, and the nagging sense that something is sitting unanswered.
An office in Melville
A professional services firm of about twenty. New-client intake comes in by email, gets copied into a spreadsheet by whoever is free, and then waits. The delay is never the work itself: it is four days of a message sitting between two people who each assumed the other had it.
The finding in a case like this is often that the bottleneck is a hand-off, not a task. The step worth building is an after-hours one that reads the intake, prepares a structured summary and a draft acknowledgement, and stops there for a person to send. What the team keeps doing: the fit decision, the scoping, the pricing, the relationship. What gets lifted off: retyping the same six fields into a spreadsheet and discovering on Thursday that Monday's message was never acknowledged.
A trades contractor off Route 110
Twelve in the field, two in the office, and quoting eats three evenings a week. The estimator knows the price the moment he has walked the job. What costs him the evening is assembling the document: pulling the rate sheet, finding the three similar jobs from last year, writing the exclusions, formatting it.
The step here drafts the quote from the approved rate sheet and comparable past jobs, fills in the standard exclusions, and then stops, because the number is still his call and always will be. What the team keeps doing: walking the job, setting the price, deciding what is in and what is out, and standing behind it. What gets lifted off: the assembly, the lookup, and the three evenings.
How the work is sequenced
First, one workflow gets chosen: the recurring one that a specific person can describe out loud in under two minutes, that happens at least weekly, and that nobody enjoys. Not the biggest problem; the most describable one. Then it gets watched end to end with the people who run it, and written down: what enters and from where, who makes each judgment call, where the work changes hands, where the sensitive material sits, and which mundane parts a machine can carry.
Before anything moves, a privacy and IT preflight documents the data classification and the approved tools, accounts, and environments for that specific engagement. Customer records, pricing, and personal information get named explicitly before any tool is discussed, not after.
The output is one recommended next step with an owner's name on it. Sometimes that step is not AI at all. A finding of "you have four spreadsheets and need to agree which one is the source of truth" is a real finding, and a consultant whose answer is always software is selling rather than diagnosing.
The first step, and what it costs
The first conversation is fifteen minutes and costs nothing. Its only job is to pick the workflow, and if the honest answer is that yours is already fine, that is what gets said.
After that, the entry engagement is the AI Assessment, a $3,000 flat engagement: one recurring workflow mapped end to end, the boundary named, and an assessment record you keep: the current-state map, a fit-and-limits read, the control points, and the recommended next step. That record is useful even if the work stops there. It is not held in a consultant's portal, it does not expire, and it does not need rTech to interpret it.
If your workflow needs steady attention rather than a one-time hand-off, the AI Concierge is a recurring retainer scoped per engagement. Workflow kits are quoted alongside an assessment when a reusable one-job workflow fits.
If you run a business in Huntington, Melville, or anywhere along Route 110, and one recurring workflow keeps costing your people their evenings, that is the one to bring.